When text information (hereinafter referred to as “document”) such as a Web page and a manuscript is browsed, the user demands to easily obtain information relating to an interesting keyword that is included in the document. For example, there is disclosed a method of presenting related information for keywords included in a region preset by the user in advance (for example, see JP-A 2009-37454 (KOKAI)). In addition, there is proposed a method of directly selecting a keyword in a Web page and retrieving related information for the selected keyword.
However, in the method disclosed in JP-A 2009-37454 (KOKAI), keyword extraction is performed for a preset screen region without determining in which region in display a user is interested. Therefore, it is impossible to perform keyword extraction in consideration of the region in which the user is interested, or in consideration of relation of the region to another region. In addition, keyword extraction for the whole page has the problem that the number of extracted keywords is large. Besides, in the case of using the method in which the user selects a keyword by manual operation, the keyword selection operation itself is complicated, in particular in small terminals.
Therefore, in the region-of-interest extraction apparatuses, it is required to estimate a keyword in which the user is interested, and estimate a region in which the user is interested at present from the document that the user is browsing, to retrieve information relating to the keyword.